Skip to contents

PRIMED: Preliminary Investigation of Meta-analytic Databases

Source: R/PRIMED.R
primed.Rd

Runs the four-step exploratory/preliminary data analysis workflow for meta-analysis of dependent effect sizes as described in Pustejovsky, Zhang, & Tipton (2026).

Usage

primed(
  data,
  es_col,
  se_col,
  study_col,
  sample_col = NULL,
  n_col = NULL,
  moderators = NULL,
  es_type = c("SMD", "correlation"),
  df_col = NULL,
  rho_values = c(0.1, 0.3, 0.5, 0.7, 0.9),
  fence_multiplier = 3
)

Arguments

data

A data frame containing the meta-analytic database.

es_col

Character. Name of the column containing effect size estimates.

se_col

Character. Name of the column containing standard errors.

study_col

Character. Name of the column identifying studies.

sample_col

Character. Name of the column identifying samples within studies. If NULL (default), assumes one sample per study.

n_col

Character. Name of the column containing total sample sizes. If NULL, sample size plots are skipped.

moderators

Character vector of column names to examine as potential moderators. If NULL (default), no moderator analysis is performed.

es_type

Character. Type of effect size: "SMD" (standardized mean difference) or "correlation". Affects how scaled SEs and weights are computed. Default is "SMD".

df_col

Character. Name of the column containing degrees of freedom (used for scaled SE calculation when es_type = "SMD"). If NULL, scaled SEs are not computed.

rho_values

Numeric vector. Assumed within-sample correlations for ISC weight calculations. Default is c(0.1, 0.3, 0.5, 0.7, 0.9).

fence_multiplier

Numeric. Multiplier of the IQR for outlier fences in effect size density plots. Default is 3 (following Tukey's conventions as described in the paper).

Value

A named list with elements:

summary

A list of summary statistics about the database.

plots

A named list of ggplot objects for each workflow step.

tables

A named list of summary tables (tibbles).

Details

The PRIMED workflow proceeds in four steps:

  1. Data structure: Counts observations at each level and describes the dependence structure (effects per sample, samples per study, sample size distributions).

  2. Moderators: Examines marginal distributions, missingness, and hierarchical (within- vs between-sample) structure of covariates.

  3. Standard errors & weights: Inspects SE distributions within samples, computes scaled SEs (for SMDs), and calculates ISC weights under varying assumed correlations.

  4. Effect size distribution: Visualises marginal and sample-level densities with outlier fences, and produces a hierarchical forest plot of dependent effect sizes.

Examples

if (FALSE) { # \dontrun{
results <- primed(
  data = my_meta_data,
  es_col = "g",
  se_col = "se",
  study_col = "study",
  sample_col = "sample_id",
  n_col = "n_total",
  moderators = c("intervention_type", "mean_age", "pct_female"),
  es_type = "SMD",
  df_col = "df"
)

# View all step-1 plots
results$plots$step1_es_per_sample
results$plots$step1_samples_per_study

# Access summary statistics
results$summary
} # }